Tuesday, June 8, 2010
Behind The Scenes
The truth about HD video capture in DSLRs
|
 The D-Movie mode of the Nikon D90, the first DSLR to offer video capture. BELOW: The Live MOS sensor of the Panasonic LUMIX DMC-G2. The Micro Four Thirds-type sensor is unique for an HD-capable camera in that it includes aspects of both CCD and CMOS sensors for reducing power consumption and increasing readout rates of data. The G2 is mirrorless and offers 720p at 30 fps, 60 fps and 50 fps (the European PAL standard). |
Binning, which combines groups of pixels into single “super-pixels,” conceivably would increase light sensitivity and speed while decreasing noise considerably, which is why Phase One’s Sensor+ medium-format technology offers the option for still capture. (Video DSLRs do produce much better image quality at higher ISOs than HD camcorders, although this is certainly due primarily to the DSLRs’ much larger image sensors and pixels.) Most of the speculation has centered on line-skipping, however, where it has been suggested that the vertical resolution is reduced by using one out of every third line to reduce resolution to the required vertical resolution. The math doesn’t quite sync up, so there still would need to be some interpolation either before or after.
HD video from still cameras is stunning, there’s no question there, but the controversy behind downsampling suggests that vertical detail has been compromised by at least a third. This would explain to a great extent why video from DSLRs doesn’t perform as well as a camcorder does when used with moiré test charts. Skew and wobble (aka the jello effect) are also frequent by-products of the nature of CMOS sensors, as image data is read out one line of pixels at a time, which causes small, but inevitable temporal delays between each line.
 |
The large pixel counts of DSLR sensors create a great demand for processing and power consumption before Live View is even introduced. DSLR image processors also have to handle many still-camera functions, including exposure, focusing, focus tracking and more. Downsampling 12- to 21-megapixel images to 1920x1080 (2 megapixels) or 1280x720 (0.9 megapixels) and compressing the data 24 to 60 times a second requires even more processing power, and only recently has DSLR processor technology achieved this. Canon’s DIGIC 4, Nikon’s latest EXPEED, Pentax’s PRIME II, Panasonic’s LUMIX Venus Engine HD II and Samsung’s DRIMe finally provide enough processing power to turn voluminous DSLR sensor output into HD video.
Selectable frame rates, a standard feature on most camcorders, allow operators to influence the look of video by alternating the rate at which the individual frames are captured; 24 fps is considered the most “film-like,” making it highly desirable when capturing video; 30 fps is the North American TV standard, and a faster rate of 60p provides smoother action. As time goes on, the processors in DSLRs are getting better and more affordable at the same time that image-processing capability is increasing, which all leads to better ways to manage video.
| Sounds Good |
| All video-capable DSLRs come with built-in microphones, and most offer input jacks that will allow you to add an external stereo mic. The internal mics are adequate for recording the ambient noise of a scene, but they also will pick up many of the sounds coming from the camera itself (autofocus, for instance). For professional video, it’s generally recommended that a professional mic be used to record concurrent sound, which means extra steps in the editing process. Many operators use dual-system sound by recording to a separate audio source while using the onboard mic concurrently. Some also offer pro XLR audio input adapters for a DSLR. |
Page 3 of 4
top
